Dr. Andreas Mueller is an Assistant Research Scientist at NYU’s Center for Data Science. We recently asked him a few questions regarding his work at CDS, and his past experience in the world of data science.
Can you tell us a bit about your educational and professional background, and what you were doing before you joined CDS?
I received my PhD in Computer Vision from the University of Bonn in Germany, and then worked on Computer Vision and Machine Learning projects at Amazon. My personal projects, however, have revolved around building open source data science application toolkits and libraries, such as Pystruct and Scikit-learn.
How did your work in computer vision factor in to your work at amazon?
There were two primary things I worked on. The first was a Computer Vision application for Amazon’s Market Place. I built the computer vision software used to compare a seller’s uploaded image to Amazon’s image standard.
The second project was using time series forecasting for product demand. The requirement was to predict the demand for a product over the coming year, in order to maintain the correct level of inventory.
What sorts of projects do you think a deep understanding of Computer Vision is particularly useful for?
It depends on the task and whether you want to do research or just work on an application. For many applications in computer vision, you can use pre-trained out-of-the-box convolutional neural networks, which often give very good results without any in-depth knowledge of Computer Vision. If you want to go deeper into the problem, you definitely need a solid background, though. Professor Yann Lecun’s class on deep learning is particularly helpful for this.
In Data Science, what are the differences you perceive in research at a company like Amazon, vs research at an Academic Institution? How was the transition from Amazon, to CDS?
The work at Amazon was primarily about solving real life problems, and so on the one hand, it is fulfilling to see the utilization of your work. On the other hand, this means you need to build a production ready application, so the project includes a lot of deployment, infrastructure and project management.
Whereas in an academic institution, your code is not the point of focus, its your research work and the papers you publish. No one actually cares about the details of the code, as long as your results were accurate and relevant.
But my current work at CDS is a mixture of both. I am free to work on whatever I want, but the software projects I develop are being used by companies like Facebook, Twitter, and Amazon, and hence need to be working on a highly practical level. I got to know that of companies like Facebook, Amazon and Twitter are already using some of my work.
How has your experience in American academia been different from when you did your Ph.D work in Germany?
In Germany, traditionally there is no coursework in PhD programs, unlike the programs in the US. Research is much more competitive in the US. Here, we have assistant professors, who are already teaching and work especially hard to gain tenure, whereas in Germany, there aren’t any assistant professors, they would already have tenure.
Can you tell us some more about your work at CDS?
I work on building out the software group, and developing tools for machine learning and data science. We are building a team with open source project experience, and we are also working on engaging domain scientists.
Of the work that you do at CDS, what feels particularly important?
Developing good functional software applications and toolkits is important to me. Even though researchers are always focusing on papers, they rely heavily on various tools and libraries for research. Unfortunately, I don’t think these projects get enough attention, even though they are such an important part of our research.
How has your overall experience been at CDS, NYU?
I like working at NYU. There are a lot of great events, talks, happenings, etc. at CDS. With my current work, I have a lot of freedom, but it’s also given me the ability to have a great impact. From David Hogg’s physics team, to David Sontag’s group, to Brian McFee’s work in Librosa, it’s really excited to be able to collaborate with people from such a diverse background.
Interview by Rishabh Jain